The Meaning of Science

Part One

What We Mean by Science

Chapter One

How Science Works

Science and Pseudoscience

There are sciences. Physics is one, chemistry another. There are also disciplines that involve the generation of knowledge and insight, but that few of us would immediately think of as sciences. History and literary studies are examples. All this is fairly uncontroversial. But there are cases where we are unsure about what counts as science, and these cases are sometimes politically and culturally explosive.

Consider the trio of economics, intelligent-design theory, and homeopathy. The only thing that unites these three endeavors is that their scientific status is regularly questioned in ways that provoke stormy debate. Is economics a science? On the one hand, like many sciences, it oozes both mathematics and authority. On the other hand it is poor at making predictions, and many of its practitioners are surprisingly blasé when it comes to finding out about how real people think and behave.¹ They would rather build models that tell us what would happen, under simplified circumstances, if people were perfectly rational. So perhaps economics is less like science, and more akin to The Lord of the Rings with equations: it is a mathematically sophisticated exploration of an invented world not much like our own.

The theory of intelligent design has been promoted by organizations like the prominent US think tank The Discovery Institute, and developed by theorists including the biochemist Michael Behe and the mathematician/philosopher William Dembski. It aims to compete with the theory of evolution as an account of how species became well adapted to their surroundings. It suggests that some organic traits are too complex to have been produced by natural selection, and that they must instead have been produced by some form of intelligent oversight: perhaps God, perhaps some other intelligent agent. The theory is positioned as a science by its adherents, but many sensible commentators think that this is merely an attempt to insert a contentious interpretation of religion into schools, and that—understood as a piece of science—the theory is hopeless.²

Mainstream doctors sometimes value homeopathic remedies, in spite of the fact that their track record of validation by large-scale clinical studies is poor. One camp says that these are quack treatments with no scientific credentials, whose apparent effectiveness derives from nothing more than the placebo effect.³ Another camp tells us that the dominant method by which scientific investigation establishes the credentials of medical interventions gives us generic wisdom regarding what works in typical circumstances for average patients, but that this approach ignores the need for doctors to prescribe what is right for a unique individual in idiosyncratic circumstances.⁴

These questions about the markers of proper science are important. They affect the power held by people whose advice can determine our financial and social well-being; they affect what our children are taught at school; they affect what forms of research our tax contributions can be used to fund and how our doctors advise that we maintain our health. These questions are also old: while today we might be concerned by the scientific status of enterprises like economics, intelligent design, and homeopathy, previous thinkers have been troubled by the scientific status of Marxism, psychoanalysis, and even evolutionary biology. What we need, it seems, is a clear account of what makes something a science and what makes something pseudoscience. What we need, it seems, is Karl Popper.

Sir Karl Popper (1902–1994)

It is still the case that if you ask a scientist to reflect on the general nature of science, you will probably be referred to the pronouncements of Karl Popper. Popper was born in Vienna in 1902, a time when Viennese cultural life was blessed with an extraordinary richness. He began attending the University of Vienna in 1918, where he exposed himself to the conspicuous intellectual movements of the moment. He became involved with left-wing politics, he adopted Marxism for a time, he listened to a lecture on relativity theory by Einstein, and he briefly served as a volunteer social worker in one of the clinics founded by psychotherapist Alfred Adler. In 1928 Popper was awarded a PhD in philosophy, and by 1934 he had published his first book, Logik der Forschung (later translated into English as The Logic of Scientific Discovery).⁵ The broad conception of scientific progress laid out in that book would remain more or less intact in Popper’s thinking until his death.

Popper—whose parents were of Jewish origin—was forced to leave Vienna in the 1930s. He moved to New Zealand, to a position at the University of Canterbury in Christchurch, where he spent nearly ten years before moving back to Europe. In 1946 he was offered a post at the London School of Economics (LSE), which he held until his retirement. The philosopher of science Donald Gillies, who first met Popper at the LSE in 1966, recently painted a lively picture of some of Popper’s idiosyncrasies:

Waiting in the lecture hall for Popper to appear was not without some amusement, because a ritual was always performed before the great man entered the door. Two of Popper’s research assistants would come into the room before him, open all the windows, and urge the audience on no account to smoke, while writing: NO SMOKING on the blackboard. Popper had indeed a very strong aversion to smoking. He claimed that he had a very severe allergy to tobacco smoke, so that inhaling even a very small quantity would make him seriously ill. When his research assistants had reported back that the zone was smoke-free, Popper would enter the room.⁶

Gillies goes on to explain that when Popper went to a specialist in allergies, the expert was unable to find any evidence of an allergy to tobacco smoke: “Popper’s comment on the result was: ‘This goes to show how backward medical science still is.’”⁷

Perhaps the high point of Popper’s reputation came in the late 1960s and early 1970s. He was knighted in 1965, and around this time a string of distinguished scientists described his work in tones of dazzled admiration. Sir Peter Medawar, a Nobel Prize winner for medicine, said simply: “I think Popper is incomparably the greatest philosopher of science that has ever been.” Sir Hermann Bondi, mathematician and cosmologist, took the view that “there is no more to science than its method, and there is no more to its method than Popper has said.”⁸

Some more of Donald Gillies’s recollections make it clear that Popper could provoke exasperation, as well as admiration. On Tuesday afternoons, the London School of Economics hosted the “Popper Seminar,” where visiting speakers were invited to present their philosophical views. In a standard academic seminar of this kind, the speaker might talk unmolested for thirty or forty minutes, before the chair invites questions from the audience. At the Popper Seminar, things were different:

Usually the speaker was allowed to talk for only about 5 to 10 minutes before he was interrupted by Popper. Popper would leap to his feet, saying that he wanted to make a comment, and then talk for 10 to 15 minutes. A typical intervention by Popper would have the following form. He would begin by summarising what the speaker had said so far. Then he would produce an argument against what the speaker had said, and he would usually conclude with a remark like: “Would you agree then that this is a fatal objection to your position?” As can be imagined such an attack would often have a very disconcerting effect on the visiting speaker.

Gillies adds: “It is easy to see that while, from Popper’s point of view, his seminar could be seen as a perfect example of ‘free criticism,’ it could have seemed to the speaker very much like a session of the committee on un-Popperian activities.”⁹

“What Is Wrong with Marxism, Psychoanalysis, and Individual Psychology?”

Popper’s basic outlook on science derived from two underlying sources of discomfort. He had grown up in a place and a time of intoxicating intellectual excitement. He recalled that “after the collapse of the Austrian Empire there had been a revolution in Austria: the air was full of revolutionary slogans and ideas, and new and often wild theories.”¹⁰ Various grand intellectual systems of exceptional ambition—Einstein’s relativity theory, Karl Marx’s theory of history, diverse psychoanalytic understandings of the mind—were in common currency. And yet, Popper felt that there was a deep difference between relativity theory, which he venerated, and (for example) psychoanalytic theory, of which he was deeply suspicious.

He set himself the task of clarifying his intuition: “What is wrong,” he asked himself, “with Marxism, psychoanalysis and individual psychology? Why are they so different from physical theories, from Newton’s theory, and especially from the theory of relativity?”¹¹ Popper’s view was that while Einstein had proposed a theory that was heroically vulnerable to destruction if experiment should show it false—and yet it had nonetheless enjoyed spectacular experimental successes—the psychoanalytic theory of mind was couched in such noncommittal terms that it was immune to experimental refutation. “I felt,” he said, “that these other theories, though posing as sciences, had in fact more in common with primitive myths than with science; that they resembled astrology rather than astronomy.”¹²

The problem with the predictions of newspaper astrology columns is not that they don’t come true: the problem is that they are formulated in such a way that they cannot but come true, and because of that they say nothing of value. My own Daily Mail horoscope for the week I write these words tells me: “You have faced more downs than ups in recent weeks, but now things are about to change. With both the Sun and Venus, planet of harmony, entering your birth sign this week, you can stop worrying about the past and start planning for the future. This is also the time to bring to the boil something that has been on the back burner for too long.”¹³ How often would we think it sensible to advise someone to “stop planning for the future and start worrying about the past”? If something has indeed been on the back burner for “too long,” doesn’t that make it trivially true that now is the time to address it? And how on earth are we supposed to quantify the relative number of “ups” and “downs” we have had over the course of weeks? It is hard to see how we can argue with any of these platitudes.

Similarly, Sigmund Freud recalled how a female patient, whom he described as “the cleverest of all my dreamers,” told him of a dream that seemed to refute his own theory of wish fulfillment. That theory says that in dreams our wishes come true:

One day I had been explaining to her that dreams are fulfilment of wishes. Next day she brought me a dream in which she was traveling down with her mother-in-law to the place in the country where they were to spend their holidays together. Now I knew that she had violently rebelled against the idea of spending the summer near her mother-in-law and that a few days earlier she had successfully avoided the propinquity she dreaded by engaging rooms in a far distant resort. And now her dream had undone the solution she wished for: was not this the sharpest possible contradiction of my theory that in dreams wishes are fulfilled?¹⁴

This woman dreamed, not of something she wanted to do, but of something she abhorred: a holiday with her mother-in-law. In spite of apparent refutation, Freud argued that his theory was intact: “The dream showed I was wrong. Thus it was her wish that I might be wrong, and her dream showed that wish fulfilled.”¹⁵ A dream that seems to jar against Freud’s theory is explained away with the argument that the woman wanted Freud to be wrong, and the dream allowed this desire to be fulfilled. It is hard not to share Popper’s discomfort in the face of examples such as these. Freud’s ability to cook up interpretations of the evidence that bring it into line with his theory hardly seems a strength of his psychoanalytic approach; instead, the elastic ability of his theory to stretch around whatever evidence may confront it seems more like a weakness.

The Problem of Induction

One set of Popper’s concerns derived from this urgent sense that we should be able to give a “criterion of demarcation’” that will tell us how to sort science from pseudoscience. The second set of concerns came instead from Popper’s deep skepticism of what philosophers call inductive inference. The eighteenth-century Scottish philosopher David Hume is usually credited with being the first to pose what we now call “the problem of induction.” To understand this problem, we first need to understand the nature of deductive—as opposed to inductive—inference.

Suppose you know that all badgers are mammals, and you know that Brock is a badger. Given these premises, you can safely conclude that Brock is a mammal. This inference is deductively valid, meaning that it is strictly impossible for the premises of the inference to be true, and the conclusion false. There is no way that we could imagine circumstances under which all badgers are mammals, Brock is a badger, and yet Brock is not a mammal. Good deductive inferences deal in certainty: their premises ensure their conclusions. Because of this, deductive inferences are often trivial or unproductive: there is a sense in which, armed with the knowledge that Brock is a badger, and that all badgers are mammals, you are simply spelling out a self-evident consequence of those pieces of information when you go on to conclude that Brock is a mammal.

Inductive inferences are different. Suppose you have invented a new drug—let’s call it Veritor—and you want to find out if it is safe. You test it on ten thousand people, and over a period of many months you do not detect adverse side effects in any of them. The people you choose to test are not all the same: you make sure you have tried the drug out on men, women, people of different ages, and people from different countries. Now suppose you ask the question: “Given that everyone tested so far has experienced no side effects, should we expect Colin, who has never taken the drug before, to experience adverse side effects?” I doubt that anyone would say we can be absolutely sure that Colin will be fine, but most people would say that it is reasonable to expect, on the basis of our extensive testing of the drug, that Colin will probably experience no adverse reaction.

Inferences of this sort are potentially far more valuable than deductive inferences, for they promise to generate important new knowledge. By looking at large, but limited, samples of people, we presume that we can make fairly reliable predictions about how other people are likely to react. Our practices of drug testing—and almost all other forms of knowledge-generation—seem to presuppose that it is reasonable to generalize in this way, via extrapolation from a limited number of observed instances. What makes this presupposition reasonable? The challenge inherited from Hume is to provide a justification for inductive inferences of this sort.

An inductive inference can be defined as any pattern of argument that we regard as reasonable, but which does not claim deductive validity. Our inference about Colin is not deductively valid, and it does not pretend to be. It does not deal in certainty, for clearly it is possible for ten thousand people to have experienced no side effects and for poor Colin to be the first to react badly. Such circumstances can easily be imagined without contradiction—perhaps Colin has an exceptionally rare genetic mutation—and it is partly because of this that we cannot be sure that Colin will be free from adverse reactions. Even so, we do take the view that our evidence, derived from testing thousands of people, makes it reasonable to conclude that Colin is unlikely to suffer adverse reactions. What makes this inductive inference reasonable?

We might try to justify our inference by appealing to further pieces of scientific research. For example, we might point out that for Colin to react in a way that is different from every one of the ten thousand individuals we tested previously, Colin would need a very unusual sort of body. We might go on to claim that it is reasonable, although not a certainty, to think that Colin’s body is typical, because human conception and development run along well-understood lines. The processes by which human bodies are typically made have been studied in painstaking detail by physiologists and developmental biologists, and this research gives us knowledge about how Colin’s body probably works, what constitutes his genetic makeup, and so forth.

This appeal to background scientific knowledge does not solve Hume’s problem. It simply reveals the depth of our reliance on inductive inference. Scientists have studied a limited number of embryonic unfoldings—in humans, other mammals, and various additional species. We assume that the processes that went into the construction of Colin were most likely similar to the processes that have been observed in the laboratory. Our inference about Colin’s constitution is based on extrapolation, and Hume’s challenge was to explain why this form of extrapolation should be thought reasonable.

The problem of induction can be put forward as a pithy dilemma: we want to know what, if anything, makes it sensible to extrapolate from a limited sample to a broader generalization. We cannot try to answer this by claiming deductive validity for our inference, for there is evidently no contradiction in the claim that our new case is freakish, and utterly unlike what we have encountered before. But if instead we try to answer our question by pointing to scientific knowledge, or even to the past successes of previous inductive inferences, it seems we are just offering yet more instances of the very extrapolations we are trying to justify. Either way, our initial challenge—what makes extrapolation reasonable—remains unanswered.¹⁶

It is time to bring our discussion of induction back to Popper. Faced with a tricky crossword puzzle, we know there must be a solution even if we aren’t quite sure what that solution is. Most philosophers—but not Popper—think of the problem of induction as a puzzle in this same sense: they have had a devilishly difficult time figuring out what the answer to Hume’s challenge is, but they are confident there must be a good answer. After all, no one gets by in day-to-day life without induction. We are all convinced that it is better to attempt to leave a room by opening a door than by walking through the wall. We are so convinced because we extrapolate from past experience of bumps, bruises, and the frustration caused by walking into solid surfaces. When our financial advisors remind us that past successes of investments may not indicate their likely future performance, we accept their warnings because we know how often healthy funds have crashed in the past. Even here, we project past patterns into the future, and we think these extrapolations are sensible.

Popper is an outlier in the debate over induction. He understood Hume to have shown that induction is a bad inferential strategy. A rational person, says Popper, is one who refuses to use inductive inference; that is, she refuses to extrapolate from past to future, from a finite number of observations to a more general theory, or from a limited number of data-points to a broader pattern. Popper’s conviction was that “theories can never be inferred from observation statements, or rationally justified by them. I found Hume’s refutation of inductive inference clear and conclusive.”¹⁷ Popper therefore set out to show how science could proceed using nothing but deductive reasoning.

Falsificationism

Popper’s philosophy of science is founded on an undeniable logical asymmetry. As we have seen, no matter how many individuals you have tested and found to respond positively to Veritor, deduction will never tell you that all people respond positively to Veritor. On the other hand, if you find just one person who responds badly to Veritor, you can conclude—with deductive certainty—that the statement “All people respond positively to Veritor” is false. If, as Popper recommends, we need to do science without appeal to inductive reasoning, then while we can never conclude reasonably that scientific generalizations are true, we can conclude that some are false, or so it seems. That is why Popper’s view is known as falsificationism.

One might think that scientists use a variety of data—from the fossil record, from DNA sequences, from the behavioral and anatomical features of plants and animals—to build a case for a more general claim like “All plants and animals are descendants of a common ancestor.” That conception of science, says Popper, is mistaken. Only science founded on induction could aim at the slow accumulation of evidence in favor of particular hypotheses, and Popper regards induction as irrational. Instead, science must proceed by a process of “conjecture and refutation”: the scientist begins by formulating a general claim about the nature of the world and then seeks to refute it by gathering data—regarding fossils, DNA, behavior, and anatomy—which, if they go the wrong way, have the potential to show decisively that our general claim about ancestry is false.

This helps us to understand Popper’s use of falsificationism to supply a “criterion of demarcation,” which pinpoints the difference between science and what Popper sometimes called “pseudoscience,” sometimes “metaphysics.” Bona fide science, he says, must be falsifiable. What makes something a genuine piece of science is its potential vulnerability to refutation. Popper was particularly impressed, for example, by the way in which Einstein’s relativity theory had laid itself open to the tribunal of experiment. As we will see in more detail a little later, Einstein’s theory made explicit predictions for the bending effect that the Sun would have on light arriving at the Earth. It thereby exposed itself to falsification if light turned out not to behave in this way. A properly scientific theory, says Popper, sticks its neck out regarding the sorts of events that it does not permit, hence regarding the sorts of potential pieces of evidence that would lead to the theory being abandoned.

Popper’s recipe has considerable intuitive appeal. Freud’s theory of the mind is written off as a piece of pseudoscience, because rather than stating in clear ways the sorts of behaviors that would lead to the theory being dropped, Freud offers slippery formulations of his commitments and slippery interpretations of his data. Likewise, the problem with astrology seems to be that its claims are stated in such intolerably vague ways that we cannot judge what it would take for the theory to be shown wrong. Things seem different with astronomy: Newton’s theory tells us precisely when to expect the arrival of a comet, and one might think that if things don’t turn out that way, so much the worse for Newton’s ideas.

The noted physicist Richard Feynman (yet another Nobel laureate) expressed a strikingly similar conception of science—surely influenced by Popper—in a lecture he gave in 1964:¹⁸

In general, we look for a new law by the following process. First, we guess it. . . . No, don’t laugh, that’s really true. Then we compute the consequences of the guess, to see what, if this is right . . . it would imply, and then we compare those computation results to nature, or . . . to experiment or experience. We compare it directly with observations to see if it works.

Feynman continued with a short summary of the falsificationist approach to scientific method:

If it disagrees with experiment, it’s wrong. In that simple statement is the key to science. It doesn’t make any difference how beautiful your guess is, it doesn’t make a difference how smart you are, who made the guess, or what his name is. If it disagrees with experiment, it’s wrong. That’s all there is to it.

Gran Sasso

In September 2011 a team of researchers announced that subatomic particles called neutrinos, sent from the CERN facility in Geneva, had been recorded traveling faster than light when their speed was measured at the Gran Sasso facility in Italy.¹⁹ Einstein’s special theory of relativity proposes an upper speed limit governing the universe: nothing travels faster than light in a vacuum. Experiment was inconsistent with Einstein’s theory. Feynman’s summary of the scientific method predicts that in spite of special relativity’s beauty, Einstein’s name, and his formidable intelligence, the results from Gran Sasso would lead to this esteemed theory being discarded.

This is not what happened. While newspapers lingered for a while on these results, most scientists felt fairly securely that the experimental results were probably flawed. They felt they were flawed partly because of their confidence in the theory those results appeared to contradict. The truth is that scientists do not throw out their theories whenever an experiment appears to contradict them. This attitude is perfectly sensible, because we are often unsure whether experiments have been conducted properly and what their true significance might be. It is perfectly rational to bet on an experiment being flawed, as opposed to putting our money on a well-tested theory being false. This observation causes no trouble at all for the practice of science, but it causes plenty of trouble for Popper’s goal of showing how science might proceed without induction.